On Tue, 2016-12-06 at 09:41 +0100, Martin Kletzander wrote:
> > When generating a new configuration for dnsmasq, add except-interface
> > for both lo and lo0: this should make sure the configuration works,
> > and is always the same for every supported OS.
> >
> > Adjust the data of networkxml2conftest to reflect the new dnsmasq
> > configuration, making networkxml2conftest pass on non-Linux OSes.
>
> Just so I know the reasoning behind this. Is this only because of the
> test suite failing? Or are we trying to fix some problem with it?
> Can't we get the interface name programmatically and then use that
> value? Not that I'd be running around adding bridges with name ==
"lo0"
> in libvirt's networks, but still... It would feel cleaner.
I think it's just because of the test case, which by the way
is still failing on FreeBSD.
I've just posted an alternative approach that, while quite
possibly not the cleanest out there, at least doesn't require
us to bend the library code proper to the test suite needs:
https://www.redhat.com/archives/libvir-list/2016-December/msg01123.html